[andyr]

I have read that back-EMF from a speaker is an "issue" in terms of one's sonic enjoyment ... ie. it's a factor which can cause problems to one's amp.

Could someone explain what back-EMF is and what causes it.

For instance, is it the fact that most cone speakers are an inductive load for an amp which causes them to produce back-EMF ... so speakers that present an almost entirely resistive load (like Maggies) will produce much less back-EMF?

Regards,

Andy

[Kevin Haskins]

Move a wire through a magnetic field and it creates a flow of current in the wire.    A voicecoil is a long wire moving in a magnetic field.   It moves due to the input signal.   The back EMF is nothing more than the resultant back voltage potential that is always exactly opposite the input signal.  

It has the effect of increasing the impedance because you don't actually have a back flow of current.  You have a potential in the opposite direction of the input signal.

This of course is a form of distortion.   How do you minimize it?  Use small voicecoils (less wire moving in the magnetic field).

[andyr]

Thanks for the explanation, Kevin.

Regards,

Andy

[Aether Audio]

andyr,

To add to Mr. Haskins comments, I would point out that the "distortion" he refers to is the result of several more complex and detailed factors.  As he quite correctly points out, the movement of a piece of wire (and even more so, a coil of it) in a magnetic field produces a voltage that is directly proportional to the amount of wire (measured by whatever graduation you prefer) moving through a magnetic field of given strength.  This is basic electromagnetic theory and the fundamental forces involved in traditional generators of electricity.

The problem in audio arises from the fact that the amplifier is sourcing a signal (let's call it the "demand signal") into a complex electro-mechanical assembly we call a loudspeaker.  For simplicity's sake, lets limit this loudspeaker to a simple cone based driver.  That driver can be broken down in it's mechanical system to three major sub-components.  The first is the mass of the cone, voice-coil, spider and surround components.  The second is the "springiness" of the suspension system (i.e., the surround and spider components). The third is the resistance to movement or "drag" of that same suspension system.

In an analogy of these three basic systems, there are direct counterparts in electrical circuits.  The mass can be thought of as possessing the quality of inductance (the electrical property of a coil of wire).  The "springiness" of the suspension components can be thought of as possessing the quality of capacitance (the electrical quality of a capacitor).  And the resistive "drag" of the suspension… as a resistor.

Depending on the values of these three electrical components and their physical arrangement/connections, they can form what we call an "oscillator."  Once an oscillator is energized, it will continue to produce current flow back and forth between the coil and the capacitor - often long after the excitation energy that started the process has stopped flowing into the circuit.

Well...this same condition can (and usually does) take place in our simple loudspeaker.  When the "demand signal" sent out from the amplifier drives the cone in a given direction, the system has the tendency to cause the cone to keep moving - even after the demand signal has told it to either stop or change directions.  This is primarily because of the mass of the cone/suspension assembly.  Just like a freight train can be hard to start and stop because of its mass, the loudspeaker will have this same problem.

If a simple "impulse" is sent out from the amplifier "telling" the cone to move in one direction only - then stop, the tendency is for the cone to keep going for a distance after the stopping point.  This quality is called "overshoot" and is the same problem you have trying to stop your car instantaneously.  Slamming on the brakes in your car will likely cause you to "skid" for a ways.  In the case of the loudspeaker though, once it has come to a stop it will then move back in the direction of the stopping point intended as programmed by the demand signal.  Here again it will often go a little further and overshoot in the oppos8ite direction.  This process will continue, back-and-forth, overshooting the correct stopping  point by less and less amounts until the cone finally stops in the proper position.  This back-and-forth process is called "ringing" and plagues all mechanical systems to some degree.  It is the ringing part that has an analogy to an electrical oscillator.

Adding resistance to the system will have a tendency to "dampen" this ringing.  If enough resistance is added then the system it will "undershoot."  This means the cone will start to slow down before it reaches it's stopping point.  If this happens then it will not be "obeying" the demand signal because it's slowing down before it's being told to.

Whether the system simply undershoots, overshoots once or rings - any aberration from the movement being called for by the demand signal is a form of distortion.  As bad as this may seem, there is a secondary side effect that can be just as negative.

Under any of the conditions outlined above, seeing that the cone/voice-coil system is no longer moving as it's being "told to" by the demand signal and because the voice-coil is moving in the magnetic field of the magnet - it becomes a "generator."  It suddenly switches from being a "driven element" into a "driving element."  Under such conditions of movement beyond that programmed by the demand signal, it generates a "back voltage" or...back emf (electro-motive force).  This back emf is now attempting to cause current to flow back into the amplifier.

This effect can either be major or minor, depending on the amplifier design.  If the amplifier uses negative feedback in its design, it will see this "new signal" on its out put terminals as an "error."  It knows what signal it generated (the demand signal) and it knows this new signal does not belong there.  So...the amplifier's error correction circuitry will try to produce a signal 180 degrees "out of phase" and of the exact same magnitude of this "error signal" - in order to cancel it out or "squelch it."  Essentially, the amplifier will try to make the error go away.  This whole process of error correction makes the amplifier work harder.  In fact, it keeps the error circuitry very busy.  If for some reason the amplifier makes a "mistake" in generating the proper error-correcting signal, it now has produced a new distortion product that was never there to begin with.  The “busier” the amplifier is making corrections; the more likely it will make mistakes.  That’s why it is important to know a little bit about a given speakers impedance and electrical phase performance.  Low impedances and large amounts of electrical phase variations can cause amplifier problems by “working it to death” – so to speak.

As you can see from the above, that's why there are many advocates of "zero-feedback" amplifier designs.  The view is that it's better to let the distortion resulting from the back emf to exist, rather than try to correct for it and make even bigger mistakes that can sound even worse.

My view is that of a compromise approach.  Make the amplifier as accurate as possible without feedback and then use just enough feedback to control back emf induced distortions.  But this is a whole other issue.

To compound the problem, most speakers have more than 1 driver and a bunch of capacitors, coils and resistors in their passive crossover networks.  Any of these have the potential to generate back emf voltages.  Some crossover designs can be such a bad combination with a certain amplifier that they can cause the amplifier to oscillate itself.  This is somewhat of a different issue but the fact is, back emf from a speaker/crossover network can reek all kinds of havoc with a given amplifier.  To be sure, the entire subject is quite complex as you start digging deeper.

Anyway, I hope the dissertation was of some help.  Sorry to run on but I needed a break from building speakers and crossovers! Trust me, they can cause grief in more than one way.

-Bob

[audiojerry]

My gosh, how on Earth can anyone design a good sounding amp or speaker with all that going on?  
I'd say that really emphasizes the importance of matching the right amp with the right speaker.

Thanks Bob, for that wonderful explanation!!!

[NealH]

Yes, great job Bob.  I've always said you should be a writer, an educator, and yes, a speaker designer too.  

When a speaker is modeled electrically, is it modeled as a second order system?

[andyr]

andyr,

To add to Mr. Haskins comments, I would point out that the "distortion" he refers to is the result of several more complex and detailed factors.  As he quite correctly points out, the movement of a piece of wire (and even more so, a coil of it) in a magnetic field produces a voltage that is directly proportional to the amount of wire (measured by whatever graduation you prefer) moving through a magnetic field of given strength.  ...

Thanks verry much, Bob, for your additional explanation.

If you could answer two more questions I have ... which lead on from what you've said, I'd be very grateful.  My speaker education would be complete!  

I have heard that "conventional" cone speakers exhibit quite a deal of inductance ... so they certainly can't be thought of as simply an "8 ohm" or "4 ohm" speaker.

In contrast, Maggies are (so I understand) almost entirely resistive - ie. they have a very small inductive component.  Does this lack of inductance imply they will show less back-EMF to the amp than a cone speaker?

Secondly, you said: "To compound the problem, most speakers have more than 1 driver and a bunch of capacitors, coils and resistors in their passive crossover networks".

Does this mean that the drivers in an active setup - which has no such passive components "in the way" between amps and drivers - will produce less back-EMF than the same multi-driver speaker when it has a passive crossover?

If so then i would've thought this was one big advantage of active crossovers?  

Regards,

Andy

[JohnR]

This of course is a form of distortion. How do you minimize it? Use small voicecoils (less wire moving in the magnetic field).

ie. create a driver with lower inductance... is there any more to it than that?

(I tend to think of back-emf as what makes spark-plugs work. It's not just the inductance but the interruption of current flow that is "interesting")

[Occam]

......
ie. create a driver with lower inductance... is there any more to it than that?

John,

Is it the inductance of the voice coil or all those dang reactive components in the (passive) crossover? Inevitably, the resulting phase angle(s) between the voltage and current of that back EMF does not Bode well for any amplifier that extends global feedback to the output stage. We are left with that fine dance between the amplifier and speaker designer. Both are desirous of 'bleeding edge' implementation, but when both do so, we often get something mucked up.

[Aether Audio]

audiojerry,

My gosh, how on Earth can anyone design a good sounding amp or speaker with all that going on?

Absolutely!  As I've said in the past, loudspeakers are horrible devices.  It's a small miracle that they work as well as they do.  But...that's actually true of all transducers (which a speaker is).  Transducers are apparatus, which convert energy in one form or "domain" into that of another.  Your eyes, a video camera detector tube (or "chip" these days), your ear, a microphone or even a thermometer are examples of transducers.  To this date, transducer technology is virtually always the weakest link in any device.  That's because the energy transformation process is extremely complex and difficult to make linear over a wide range of frequencies and amplitudes.

With loudspeakers, microphones and ears, it's all the more complex.  This is because there's actually two transformations taking place.  First there's the transition from the acoustic domain to the mechanical domain, then there's the transformation from the mechanical to the electrical domain.  That's a lot to ask from any transducer - and that's why they're so difficult to build and optimize.  So yes, your comment is pretty right-on.

audiojerry & rnhood,

Thanks for the kind comments!  I may be gifted to some degree in being able to explain things - for whatever that's worth.  But the bigger question is:  Do I know what the heck I'm talking about?!!! I hope so, but I don't expect anybody to just take my words as "gospel."  I would advise anybody to do your homework and check other sources.  Heck, you may be able to straighten me out when I'm off in "la-la land" sometime.

andyr,

Does this lack of inductance imply they will show less back-EMF to the amp than a cone speaker?

Well...this is actually more complex than it might seem.  The fact that the Magnapans use magnets, implies there must be some small amount of inductance, otherwise they wouldn't work.  In their case the "voice-coil" is distributed across the entire surface of the membrane so the actual inductance value is rather small.  This can be both good and bad.

Seeing that they use a large amount of magnet material, essentially what they're doing is trading coil length for total magnetic flux.  Remember: "Force Factor" is a function of "BXL."  "B" is the symbol used to signify total magnetic flux and "L" is used to signify the length of wire passing through that flux field.  To achieve the same Force Factor, if you decrease "L" (as in the Maggies) you have to increase "B."  That's why they have to have all that magnetic material behind the membrane.

Now, even though there's a small amount of inductance, there's a large membrane area that will "modulate" that inductance.  That means small amounts of undershoot, overshoot and ringing (as described in my previous post) will still be transformed in the reverse direction into back emf voltages.  Sorry, there's no "free lunch."

In this case, due to the lower inductance the Maggies may be a somewhat easier load on the amplifier, but they still have the same problem with generating distortion products.  In fact, it can even be worse than a conventional cone speaker. It's all in the fundamentals of the "stretched membrane" technology on which they are based.

As an example:  Imagine a stone dropped in the middle of a calm-as-glass pond.  It's easy to envision the ripples traveling outward from the center ad infinitum...and never to return.  That's sort of how the stretched membrane technology is supposed to work - but there's a problem.  Imagine the same stone dropped in a medium sized tub of water.  The ripples go out real nice - until they hit the walls of the tub.  Then they reflect and head back towards the center.  In just a second or so the pattern of the ripples becomes incredibly complex as they pass over one another.  Depending on the size of the original "plop" of the stone and the size of the tub, in very short order the complex pattern of ripples will completely mask the point where the wave started from.

In a stretched membrane device, the same thing can happen.  The terminating point is where the original wave reaches the edges of where the membrane is supported.  If the wave reflects at those edges it will ripple back through the membrane and produce a very complex pattern of distortion products.  Essentially ALL reflected energy is distortion and if you could visually see it, it would be very ugly.

Designers of these devices work very hard to balance the ability of the membrane to move at all with finding a way to add damping to that supporting edge so as to minimize any reflections.  It's a very sensitive balancing act and by it's very nature always results in some compromises being made.

To make the whole scenario even more complex, if an amplifier is used with a high damping factor, it will try to correct for these errors through canceling the back emf resulting from the distortion ripples in the membrane.  Problem is:  Seeing that there is less inductance, some of the lower amplitude ripples will produce such small back emf voltages, that the amplifier barely "sees" them.  This is due to whatever resistance there is between the membrane's electrical terminals and the amplifier's output terminals - i.e., speaker wire resistance.

All speakers have this same problem, regardless of design.  The issue is that as the back emf voltage gets smaller, their ratio in magnitude to the resistance of the speaker wire becomes greater.  At some point the amplifier's correction circuitry barely "sees" them and has its ability to reduce or eliminate them reduced as well.  It is highly dependant on the amplifier's "small-signal dynamic linearity."  One of the measure of this is the amplifier's output impedance or inversely stated...damping factor.  The higher its damping factor, the more it will be able to correct for these problems.

So it would seem a high damping factor amplifier would be a good mate for a stretched membrane speaker.  But then, they all use large amounts of negative feedback to achieve that high degree of damping.  We all know that territory is fraught with hazards.  That's not to say excellent sounding amplifiers using large amounts of feedback don't exist.  I know of three designs personally.  They are: a Crown Macro-Reference, a Belles Reference 350 and the Nuforce Ref 9's - all excellent sounding designs that have very high damping factors.  Are any of these good mates for Maggies et.al?  Can't say, never heard them.  But if I owned a pair of Magnapans or any other stretched membrane speaker, I'd sure give those amps or something like them a try.

Well, there you go…another dissertation.  I got to get to work!  I enjoy helping you guys though and whenever I can, I’ll be glad to answer whatever questions I can.  As a little side note, what with all the hoopla about waveguides over at the VMPS circle, I’d be glad to get into the discussion of what a waveguide is and what it’s supposed to do.  I think BC is stretching the definition of such devices quite a bit in his new patent and there’s a few others around here that agree with me.  Not to beat him up though; as long as the thing does what he claims I guess he can call it whatever he wants.  It’s all a matter of semantics but I’d hate to see folks left even more confused as a result of stretching the definition of the waveguide.  We need more clarity in this hobby – not less.

Take care,
-Bob

[art]

Wires can be arranged in a serpentine manner to reduce, if not cancel out, the inductance.

Most of my customers have Maggies, as do I. That may give you a clue as to how they work with planars.

(Another reason may be that planars tend to be less sensitive to the warts that genre of amps may have. Another topic for another time.)

Pat

[Dan Banquer]

"Is it the inductance of the voice coil or all those dang reactive components in the (passive) crossover? Inevitably, the resulting phase angle(s) between the voltage and current of that back EMF does not Bode well for any amplifier that extends global feedback to the output stage. We are left with that fine dance between the amplifier and speaker designer. Both are desirous of 'bleeding edge' implementation, but when both do so, we often get something mucked up."

When an amplifier drives a loudpeaker with a crossover the back emf from the woofer will "hit" the inductor of the crossover low pass filter first which will isolate this from the amplifier. When driving a loudspeaker directly I would recommend the same technique that is used when amplifiers drive motors. In that application a series choke is used between the amplifier and motor. For driving a loudspeaker directly I would recommend the same, especially for woofers.
So to any of you who maybe using an active crossover this technique may well improve performance.
Hope this helps;
           d.b.

[Kevin Haskins]

This of course is a form of distortion. How do you minimize it? Use small voicecoils (less wire moving in the magnetic field).

ie. create a driver with lower inductance... is there any more to it than that?

(I tend to think of back-emf as what makes spark-plugs work. It's not just the inductance but the interruption of current flow that is "interesting")

Ha.... it is a function of total B also.   Strength of the magnetic field, amount of wire you have moving through it and the distance with which you move that wire.   Making the wire shorter gives you less inductance but less motor force also.   With XBL^2 we typically increase B a smooch but what is more important is we extend it to get a wide flat BL curve.

The back EMF isn't the largest source of distortion.   Non-linear BL is a bigger factor.  Non-linear suspension behavior is about 10-20% of a driver's total distortion.   The motor is most of the rest.   This entire subject is a really dumbed down discussion of the dynamics at work.  

Dr. Klippel is doing the best analysis of driver large signal behavior.   There is plenty of good reading on his site and in his publications.    Go check out his site if you want to do more reading.   I've been though his papers two or three times and I'm still going back for more.   Wiggins is my interpreter when I get to something I don't understand.   It's pretty handy having a guy around who designs drivers from scratch to help you through the details.

[andyr]

Thanks, Bob,

Your comments are printed and in my file for future reference!  

Regards,

Andy

[andyr]

"When driving a loudspeaker directly I would recommend the same technique that is used when amplifiers drive motors. In that application a series choke is used between the amplifier and motor. For driving a loudspeaker directly I would recommend the same, especially for woofers.
So to any of you who maybe using an active crossover this technique may well improve performance....

Dan,

That's a very interesting idea ... am I correct in the following calculation?  For the inductor used for a 6dB LP roll off:
* at 1,000Hz for an 8 ohm driver it is 1.25mH
* at 3,400Hz and a 4 ohm driver it becomes 0.18mH?

I chose 3,400Hz as this is a decade away from the LP crossover point of my active crossover ... so I assume it is far enough away not to "interfere" with the active crossover.

However, I thought one of the "advantages" of an active crossover vs. a passive, was that the amp was directly in control of the driver - with no "looseness" caused by having a series inductor "in the way"??  

Also, the inductor will have an inevitable DCR which has no "positive" effect?

Can you elucidate why the advantage of dissipating back-EMF with an inductor outweighs these two "negatives"?

Thanks,

Andy

[Russell Dawkins]

Kevin,
could you point me towards Dr Klippel's site? Did a Google on Dr Klippel and everything was about arthritis!

[Kevin Haskins]

Kevin,
could you point me towards Dr Klippel's site? Did a Google on Dr Klippel and everything was about arthritis!

Sure... that may be useful information in a few years.

http://www.klippel.de/

[Dan Banquer]

Andyr: I think you are going to have to some experimenting here on your own. As I pointed out in my earlier post this "may" help. It also may not.
I have not to date had to deal with this issue. The issue of wire resistance from a choke  can be easily offset by either using an air core with a 14 AWG wire or one of the heavy duty Erse chokes with with really big core. If you maintain a damping factor greater than 50 you should be O.K.
As to the values of inductance that is going to be what you initially calculate and then you will probably revise that after test.  
At this point I can only say test, test, test, and listen. I suspect that your testing will tell you if back emf is a real issue or not in this application, because their appears to be some dispute as to how much of a factor this is.
           d.b.

Dan,

That's a very interesting idea ... am I correct in the following calculation?  For the inductor used for a 6dB LP roll off:
* at 1,000Hz for an 8 ohm driver it is 1.25mH
* at 3,400Hz and a 4 ohm driver it becomes 0.18mH?

I chose 3,400Hz as this is a decade away from the LP crossover point of my active crossover ... so I assume it is far enough away not to "interfere" with the active crossover.

However, I thought one of the "advantages" of an active crossover vs. a passive, was that the amp w ...

[andyr]

Andyr: I think you are going to have to some experimenting here on your own. As I pointed out in my earlier post this "may" help. It also may not.
I have not to date had to deal with this issue. The issue of wire resistance from a choke  can be easily offset by either using an air core with a 14 AWG wire or one of the heavy duty Erse chokes with with really big core. If you maintain a damping factor greater than 50 you should be O.K.
As to the values of inductance that is going to be what you initially  ...

Thanks, Dan,

I will try it at some stage but it's actually not a cheap experiment!  Using 0.2mH 8g North Creek inductors (which have the lowest DCR ... 0.02ohms) it would be US$35 or 40 each (their web price table seems to have a misalignment so it's hard to work out the correct price!).  Plus shipping overseas!

But how do I know "if back emf is a real issue or not"??     What factors do I listen for?

Regards,

Andy

[Dan Banquer]

I think a standard 14 AWG air core will do fine here. Remember that your voicecoil resistance is typically between 5 & 6 ohms.  Theoretically back EMF is an opposing voltage that gets thrown back at the amplifer in this application. If this is real then one would think there would be a lessening of dynamics. You may wish to do some testing of "transients" for this, and check for differences in peak levels to see if there is any.
As far as listening is concerned check for a level differences of dynamic peaks in music. Take your time here, be sure to use the same levels  and do NOT make snap decisions.  Go back and forth a few times to be  sure you percieive audibilty or not as the case may be. You just may find that the insertion  loss of the inductors may be more of an issue than back emf.
                   d.b.